AI Hiring Tools Are Screening Out Disabled Applicants. Workday Is Just the Start.

The Mobley v. Workday lawsuit isn't primarily a story about one HR platform. It's a signal that algorithmic hiring has created a structural barrier to employment for disabled people at industrial scale — and most employers using these tools have no idea what's happening inside them.

Workday processed nearly one million job applications per day in 2024. If the lawsuit's core disparate impact claims hold, that's not a software bug. That's a civil rights problem operating at extraordinary volume.

What the Lawsuit Actually Argues

U.S. District Judge Rita F. Lin dismissed the intentional discrimination claims in Mobley v. Workday (opens in new window), but she kept the disparate impact claims alive. That distinction matters enormously.

Disparate impact doesn't require proving that Workday intended to screen out disabled applicants. It requires showing that a facially neutral practice — algorithmic screening — produces discriminatory outcomes for a protected class. Under Title I of the Americans with Disabilities Act (opens in new window), employment discrimination doesn't have to be deliberate to be unlawful. The legal standard is effect, not intent.

This is how algorithmic bias typically operates. As Baltimore attorney Anthony May explained in the source reporting, AI systems trained on historical hiring data will replicate the patterns embedded in that data. If an organization's executive ranks were built during more exclusionary periods — and most were — the algorithm learns to favor candidates who resemble those executives. Disability status, gaps in employment history, non-linear career paths: these become negative signals, even if the system never explicitly evaluates disability as a category.

Workday's response — that its tools "look only at job qualifications, not protected traits" — reflects either a genuine misunderstanding of how algorithmic discrimination works, or a deliberate rhetorical sidestep. Protected traits don't need to be explicit inputs when proxy variables can do the same work.

The Employer Liability Question No One Is Asking Loudly Enough

Here's the analysis that should be keeping HR directors and general counsel awake: Workday isn't the only defendant that matters.

Employers who deploy these tools bear independent obligations under the ADA. The EEOC's 2023 guidance on AI and algorithmic fairness (opens in new window) makes clear that employers cannot outsource their discrimination liability to a vendor. If an AI screening tool produces discriminatory outcomes, the employer using it is responsible — regardless of what the vendor's terms of service say.

This creates a specific, urgent obligation for organizations like Johns Hopkins, which is currently transitioning its entire hiring infrastructure to Workday and has announced plans to use the platform to "unify the entire hiring process" for its health system. Hopkins is the largest private employer in Maryland. The scale of that deployment, combined with the pending litigation, creates a compliance posture that deserves serious scrutiny before the system goes live.

Baltimore City and Baltimore County, notably, both confirmed they use Workday but have not activated its AI screening features. That's a reasonable interim position. The University of Maryland system made the same call. These organizations aren't avoiding technology — they're exercising the human oversight that Title I compliance actually requires.

What Employers Need to Evaluate Right Now

The practical question for any organization using AI-assisted hiring isn't whether to trust the vendor's responsible AI statements. It's whether you can independently verify what the tool is doing and document that verification.

| Evaluation Area | What to Examine | Primary Authority | |---|---|---| | Disparate impact analysis | Does the tool produce statistically different outcomes by disability status, age, or race? | 42 U.S.C. § 12112 (opens in new window) (ADA Title I) | | Reasonable accommodation in screening | Can applicants with disabilities request modifications to AI screening processes? | 29 CFR Part 1630 (opens in new window) | | Vendor audit rights | Does your contract give you access to bias testing data? | Contract review required | | Human override capability | Can a human reviewer intervene when AI flags a candidate? | EEOC AI guidance | | Documentation of job-relatedness | Can you demonstrate that screening criteria are genuinely job-related? | Uniform Guidelines on Employee Selection Procedures (opens in new window) |

This table isn't exhaustive, but it maps the minimum analytical framework any employer should apply before activating AI screening features. "The vendor tested it" is not a defensible answer if the EEOC or a plaintiff's attorney asks what you did.

The Deeper Pattern: Structural Barriers Built at Scale

What makes this case significant beyond its immediate facts is what it reveals about how accessibility barriers are now constructed. Physical barriers — a missing ramp, an inaccessible entrance — are visible. They can be photographed, measured, remediated. Algorithmic barriers are invisible to the people they affect.

A disabled applicant screened out by Workday's AI receives a rejection. They don't receive an explanation that an algorithm scored their application negatively because their employment history included gaps consistent with disability-related leave, or because their educational path didn't match the proxy variables the system associates with success. They simply don't get the interview.

This opacity is itself a civil rights problem. The compliance framework challenges that organizations face in traditional accessibility work are significant — but at least the barriers are knowable. Algorithmic screening creates discrimination that is structurally difficult to detect, challenge, or remediate without vendor cooperation.

The settlement dynamics here will also be worth watching. Research on how legal victories can create compliance failures suggests that even favorable outcomes in cases like this often produce surface-level remediation rather than genuine systemic change. A settlement that requires Workday to modify its testing protocols without requiring employers to independently audit outcomes would be a procedural win that leaves the underlying problem intact.

What This Means for Employers in Practice

The "canary in the coal mine" framing from attorney Anthony May is analytically accurate, but it understates the urgency for employers already using these tools. The canary has already stopped singing. The question is what you're going to do about the air quality.

Specific steps employers should take now, regardless of how Mobley v. Workday resolves:

• Request your vendor's bias audit data — specifically, outcomes disaggregated by disability status, age, and race. If the vendor won't provide this, that's material information.
• Review your reasonable accommodation process for the screening stage itself — not just post-hire. Can applicants request human review of AI-flagged applications?
• Audit your job requirements for criteria that may screen out disabled applicants without genuine job-relatedness justification.
• Consult your contract for indemnification provisions and audit rights before the next renewal cycle.
• Document your oversight process — what human review exists, at what stage, and who is responsible.

The Southwest ADA Center (opens in new window) and regional ADA centers offer technical assistance on employment discrimination questions that can help organizations think through these obligations without waiting for litigation to clarify them.

Workday processes nearly a million applications a day. If its AI screening tools are producing discriminatory outcomes, the aggregate harm to disabled job seekers is not a compliance footnote — it's a significant and ongoing civil rights injury. Employers who deploy these tools share responsibility for that outcome. The litigation will eventually answer the legal questions. The practical question — what are you doing right now to ensure your hiring process doesn't exclude qualified disabled applicants — doesn't wait for the verdict.